RustLatam 2019 - Without Boats: Zero-Cost Async IO

What people said

sgrove · 2019-05 · Strong recommendation

The nuances in this case went well beyond what I was expecting - it's actually more than just syntax, this heavily crosses over into semantics. It's been a very impressive process, and I was surprised at the power of the solution they landed on. - Rust has to find a syntax that can work with different underlying libraries (Rust can compile to lots of different targets, some will have capability to accommodate larger libraries than others) - The abstraction can't incur undue overhead (callback-based solutions may cause allocations that are hard to keep under control and will split the language if it spreads too much) - The syntax should be pleasant to write for at least the majority of people (modeling your entire application as a state machine around async computation is pretty unpleasant) - The compiler has to be able to give helpful error messages (this was a problem early on I think) - Language authors had to think about how the borrow checker would work across async points, and whether to invest in changing the language to make that possible. - and so on and so on I really loved this talk by Without Boats (the author of the friendly article) going over all this stuff, really gave me a good sense of what they were up against [VIDEO]

Read the original HN comment ↗

xiphias2 · 2020-04 · Recommendation

I was referring to the fact that it is the only async implementation that I know of that compiles to a state machine that doesn't need to allocate new memory for each continuation. I think it will translate to extremely fast web services in practice. Pinning is of course part of the elegant solution. There are some great videos about it: [LINK] [VIDEO]

Read the original HN comment ↗

teddyc · 2019-12 · Recommendation

I thought this talk about Async in Rust was pretty cool (RustLatam 2019 - Without Boats: Zero-Cost Async IO): [VIDEO]

Read the original HN comment ↗